On 27/10/11 13:27 -0400, Chris Alfonso wrote:
The current implementation will push each target image to the same
provider
using the same provider account.
---
As discussed in irc, NACK. So we have a history of why, I'll attempt
to repeat the issue here for the list. I did run this successfully
pushing to ec2, fwiw, but the problem is, as this is currently
implemented, it will not work on anything but snapshot style
providers. The reason for this is that the call to factory to
initiate a build returns ~immediately. This does _not_ mean the build
is complete, just that the request to initiate one was successful. In
the case of upload style clouds, the build will take at least as long
as the push step for ec2 (which really is where the work of the build
happens for that provider).
The net effect of this is that if you attempt to run this command on
an upload style cloud, you will be greeted with an error when it
attempts to push a build which does not yet exist in iwhd. Note that
this particular problem is also not solved by our request to put
placeholder objects into iwhd immediately, unless factory were to
_itself_ use the build status to determine if it could initiate a
push. Even in that case, we would either get an error back that the
build could not yet be pushed, or factory would have to know to queue
up the push once the build was actualy complete (this is what I was
trying to say in my reply to the original thread, and why I was
suggesting this might not be a simple path).
-j